How to deliver tech-driven AML solutions without failing: A Kyckr Webinar

Key Takeaways:

• Stop relying on legacy systems and manual processes
• Do not look for the ‘unicorn solution’ from one vendor; greater collaboration is key
• Avoid the siloed approach when looking to implement new programmes that are designed to tackle AML - involve the tech team from the beginning

The UK financial services sector is spending £22,000 every hour fighting fraud, according to recent research by LexisNexis. 

In the fight against financial crime and with this much capital at stake, it makes sense to utilise any recent developments in technology and automation. But financial organisations are still having a hard time implementing these technologies and often struggle while trying to deliver new and tech-driven AML solutions. 

We recently held a webinar with our Chief Operating Officer, Steve Lamb, Mike Harris of FCC Consulting, and Oli Platt of NayaOne to learn why.

‍Keeping up with the times

From weekly conversations with financial organisations, Kyckr has seen that most are now in the midst of a long-term digital transformation of their compliance processes. Things are finally moving in the right direction. 

"When I joined Kyckr about three and a half years ago, lots of firms still had very manual and paper-based processes. In the last five years, there has been a clear shift toward digital, with big financial institutions and regulated entities starting to look at – and embrace – technology, using it to automate what is currently a consistently manual and costly process," said Steve Lamb, our COO. 

But despite this, many firms still rely on their legacy systems and the manual reviews in their AML processes and spreadsheets - sometimes still cutting and pasting data from multiple sources. And while they are committed to improving their AML strategy, they often struggle to keep up with the latest developments, and the delivery of tech-driven solutions has not been unproblematic. 

"If we look at a subset of AML, Know Your Customer (KYC) is an excellent example. A recent survey found that three out of four firms rated their level of digital KYC sophistication as poor or mediocre. On a scale from fully manual to fully automated, a quarter scored themselves at four out of 10 or less,” continued Steve.

So while the ambition is there to automate fully, the lofty goals firms ambitiously set out rarely come to full fruition. But why is that? Why do so many firms fail to reach their automation objectives, what are the most common pitfalls, and how can these be avoided?

‍'Dirty' data

“Throughout the history of computing and technology solutions, data is the oil that makes it all work. Technology can only deliver the benefits if it's accurate, up to date, and high quality,” said Mike Harris to introduce the continued importance of data.

When investing time and money in new tech, it is essential to look at the data you intend to feed it. While the front end of these solutions tends to be the standout element in any transformation, data is what these solutions live or die by. 

Even if you have a great platform, it still needs to be fed by great data. Often overlooked in the excitement of new tools such as OpenAI, this will impact the delivery.

“We're seeing a huge amount of work at the moment into actually standardising some of that data collection,” said Oli Platt.

“People are making sure that there's really high-quality data throughout the institutions.”

‍Unrealistic goals

Large transformation projects have big goals like 'automating the entire AML process' or 'driving a 90% reduction in manual handling time'. While inspiring and commendable, these sweeping goals are often unachievable. The problem is that companies are focusing on doing too much at once. 

According to Oli Platt, large firms often have very outlandish KPIs to measure success by. And despite having similar KPIs,  tech providers will stick to ones they know they can deliver. The sweet spot is somewhere in the middle. 

“We spend a lot of our time working with these financial institutions and tech providers and helping them aim for something realistic. The key is bringing those pieces together, having defined KPIs, having a lot of motivation, and being able to test tech quickly and evaluate if we think it will be successful.”  

There are big gains to be found even in the smaller sub-processes. It helps to break these down into a subset of mini programmes or projects and identify where the quick wins are. If companies can pocket those wins and start generating automation on a smaller scale, this will often cascade into other parts of the process. 

Another unrealistic goal is the ambition to find a one-size-fits-all unicorn solution to all the company's automation requirements. 

“I'm quite inherently sceptical about claims that there's one vendor that can do it all, that can automate every part of your process. These are just too complex. There are too many moving parts,” said Steve.

Instead, organisations should partner up, use APIs, and collaborate. They should choose the best parts from specialist vendors and organise them into a combined solution that works for them. 

‍Avoid the siloed approach

There is also a tendency to leave the tech team out of the process until the implementation stage. But only engaging the tech experts at the end of the process means implementation often runs into trouble.

“Where I've seen these deployments go wrong is where let’s say, the Head of Financial Crime has built a set of requirements. They've engaged their procurement team, they've engaged their legal team, they've got executive sign-off, they've had the procurement contracts signed, and only at that point is it chucked over the fence to the technology team.” 

According to Steve, organisations looking to incorporate a technology-focused programme should instead involve their tech stakeholders from the off. And to do that, they need at least one tech person – or ideally a team of tech people – embedded within the function who can help craft the requirements. Thus these programmes will be more likely to lead to successful delivery. 

 Technological advancements are taking centre stage in how businesses from all walks of life operate, and financial services are no different. Knowing how to implement effective tech-driven anti-money laundering solutions is a critical area in this sector, and it will be a vital component in determining the success of your AML strategy.

‍A quick glance at AI 

A hot topic across most industries this year is how artificial intelligence will impact the sector. Where are we in terms of the adoption of AI-type technologies and AI machine learning? And where are the best opportunities to improve process efficiency? 

According to Oli Platt, most people are not sure yet. 

“I’m not convinced that it’s wholeheartedly being embraced, accepted and pushed forward as of yet. We are seeing an increasing interest in ChatGPT, where the possibilities of large language models are now at the forefront of people’s minds.”

Indeed, while AI technology has potential to bring about many merits for financial institutions, most still have many questions on the topic.  

“We’re seeing a lot of financial institutions go ‘Okay, well, could we deploy our own?’ And if we could train it on our own data, what would that look like? Could that help us? Could we bring in vendors who would help us deploy these models and begin to embrace big data? 

Although all three panellists were somewhat cautious considering the benefits of AI on AML, they also agreed that in the right hands, it could be a fantastic tool – not only to assist financial institutions in their tech-driven AML search – but for all manner of things like data compliance, KYC and UBO checks, and helping with customer onboarding through faster and safer means. 

With AI and machine learning trained on large data sets, Steve said he'd be very cautious of anything where he didn’t know the underlying data set or the quality of the information it's being fed on. He also said we should look to augment existing processes with artificial intelligence rather than replacing them.

“Can I see any horizon where we feed an individual name or a business name into an AI model, and it spits out a risk rating? No, I think we're years away from that for a number of different reasons,” remarked Steve.